18 results about "Task oriented" patented technology

The invention discloses a task-oriented grabbing method and system based on cross-level constraint reasoning, and belongs to the technical field of robot grabbing control. The method comprises the following steps: aligning a text instruction with an input image through a mask alignment module, generating a target area mask by utilizing SAM-Clip, and generating a target point cloud by taking the target area mask as spatial prior; and further analyzing an internal physical structure of the target by using VLM, guiding to generate a 6-DOF grabbing attitude, performing collision detection and quality scoring by combining high-level function and bottom-level geometric prior, and outputting an optimal grabbing attitude. According to the method, the problems that task understanding and scene perception are disjointed, and the grabbing posture lacks constraint are solved, and the grabbing success rate and the intelligent level of the robot in the open environment are remarkably improved.

The invention relates to the technical field of adaptive control systems, in particular to a sofa posture adaptive adjustment system based on multi-modal sensing fusion, which comprises a baseline modeling module, a monitoring diagnosis module and an intervention decision module. According to the invention, a personalized dynamic characteristic parameter baseline is established for a user through the baseline modeling module, and unconscious attitude deviation and conscious task-oriented actions are accurately distinguished by a micro intention recognition program unit in the monitoring and diagnosis module; and the intervention decision-making module starts adaptive deviation suppression adjustment only when the former is identified. According to the invention, the technical problem that the existing control system cannot distinguish the state change reasons to cause frequent false triggering of the adjusting action is solved, and personalized and intelligent closed-loop control under the condition of avoiding invalid interference is realized.

The invention provides a virtual user simulation method and system for dialogue testing, and the method comprises the steps: obtaining a plurality of pieces of effective multi-round dialogue data from a to-be-tested system, and carrying out the multi-label labeling, so as to construct a basic sample data set containing user portrait labels, the key sample data set further comprises user emotion tags; constructing a group optimization strategy model, performing cold start training based on the basic sample data set, and performing intensive training by using a multi-dimensional composite reward function based on the key sample data set to obtain a dialogue generation model with an emotion cognition-verbal skill expression decoupling architecture; and taking a historical round of dialogue of the to-be-tested system aiming at the current interaction as input of a dialogue generation model, and generating a next round of user reply consistent with the user emotion and the user portrait. According to the invention, deep, efficient and automatic adversarial testing is carried out on the task-oriented dialogue system, so that system defects which are difficult to find by a traditional testing method are effectively exposed.

The invention relates to a task-oriented semantic communication-oriented adaptive source channel coding system, which belongs to the field of wireless communication, and comprises a sending end for extracting a semantic feature map for evaluation from input data through a feature extractor initialized based on a pre-training large model, executing nested quantization and arithmetic coding based on a calculated importance score, and sending the semantic feature map to a task-oriented semantic communication-oriented adaptive source channel coding system; generating a scalable semantic bit stream with importance levels; the physical layer is used for carrying out reordering and grouping according to importance levels of bit streams, a channel encoder is combined with an interleaver to realize unequal error protection of semantic matching, and encoded bits are mapped to sign bits with different fault-tolerant capabilities in a constellation diagram through division modulation; and a receiving end: recovering the semantic feature map through arithmetic decoding and nested inverse quantization. According to the invention, on the premise of keeping compatibility with a digital communication protocol stack, through distortion-driven joint rate self-adaption, the transmission effectiveness for machine tasks and human vision is obviously improved under a limited bandwidth.

The invention provides an intelligent report automatic generation method and system based on a multi-modal large model, and relates to the technical field of artificial intelligence, and the method comprises the steps: firstly receiving a report generation demand text, and deconstructing a demand semantic deconstruction result; constructing a multi-modal information association mapping rule based on the demand semantic deconstruction result, and obtaining and associating multi-modal information to obtain an association set; executing a task-oriented adaptive processing generation report generation task instruction; calling a large model to perform content conversion and integration on the association set according to the report generation task instruction to generate an initial report text; and finally, performing text line structure and semantic linkage optimization on the initial report to obtain a final intelligent report. According to the method, the demand can be accurately understood, the multi-modal information is fully utilized, and the high-quality intelligent report meeting the requirement is generated.

The invention provides an operation video behavior triple recognition method based on multi-task query, and the method comprises the steps: extracting the image features of a current video frame; constructing a plurality of groups of task-oriented learnable query vectors which respectively correspond to three sub-tasks; based on an original prediction vector of a previous frame of main task, screening a highest confidence vector and converting the highest confidence vector into a prior query vector; fusing the prior query vector with the current frame subtask query vector to obtain a previous frame guided subtask query vector; through a multi-head attention mechanism, enabling the subtask query vectors guided by the previous frame to interact with the image features of the current frame, and generating task specificity vectors; and inputting the task specificity vectors into a decoder and a classification head to obtain classification results of the main task and the sub-tasks. Therefore, the operation video triple recognition precision and the time sequence stability are improved, class imbalance is relieved, the model interpretability is enhanced, and clinical requirements are met.

The invention proposes a novel method for developing a TOD model by representing input data as function calls and utilizing LLM as the foundational model to train multi-tasks via finetuning LLM with instruction. Furthermore, to implement the proposed model on LLMs with moderate sizes (fewer than 10 billion parameters), the invention also presents a finetuning LLM method to enhance the capability of these LLMs in terms of handling function-calling tasks. Finally, an effective training strategy with customized loss functions for each specific task is presented to optimize the training process. Experimental results on the standard MultiWOZ 2.2 dataset demonstrate the superior performance of the proposed method compared to existing approaches in this field of research.

The utility model discloses a stroke upper limb treatment device based on task-oriented training, which comprises a pair of bases, a pair of lower supporting rods arranged on the bases respectively, n-shaped supporting frames arranged on the lower supporting rods, a lower connecting rod transversely arranged between the two lower supporting rods, and an upper connecting rod transversely arranged between the n-shaped supporting frames, the two lower ends of the n-shaped supporting frame are connected with the upper ends of the two lower supporting rods through rotating joints so that the n-shaped supporting frame can rotate front and back, and upper limb fixing accessories are further arranged on the n-shaped supporting frame. The device is simple in structure and convenient to use, and through the n-shaped supporting frame capable of swinging front and back and different upper limb fixing accessories, the requirement for treating the upper limbs according to training matched with the current functional states of different stroke patients can be met.

The application discloses a kind of three-layer component architecture and scene constraint hybrid open multi-agent system and method, wherein the method comprises: by preparing component group configuration target, meaning, signal and prior shared understanding, form scalable initial semantic basis;Through interactive component group, the interactive task of single focusing on single concept attribute is executed, and demand expression, scheme feedback and consensus verification are completed under the rule of physical concise communication;The purpose of the application is to solve the problem that in the prior art, in the hybrid open multi-agent system containing human and artificial agent, due to the ambiguity of human communication mode, the limited cognitive load, the sensitive interaction cost, and the system needs to support dynamic agent composition and unpredictable task scene, the traditional fixed communication protocol and formal semantic alignment mechanism are difficult to adapt, and then it is impossible to efficiently establish shared task-oriented understanding under the double constraints of "less assumption and less interaction".

The application discloses a kind of based on load perception's GPU resource dynamic allocation method and system, it is related to data processing technical field.The method includes: on task management platform, execute first task interpretation and mapping, generate first task atlas layer, execute multi-dimensional GPU resource state mapping, determine second resource atlas layer;First task atlas layer and second resource atlas layer are migrated to container orchestration module, migrate task intelligent agent to container orchestration module, execute task and resource allocation game, determine resource allocation strategy;Task-oriented resource allocation scheduling management is carried out to GPU resource.Solve the technical problem that GPU resource allocation lacks load perception in prior art, cannot dynamically adapt to task change, leading to low resource utilization and unreasonable allocation, achieve according to real-time load dynamic accurate allocation of GPU resource, improve resource utilization, realize the technical effect of reasonable and efficient task and resource matching.

The invention discloses a synthetic primitive-driven expert-guided multi-round dialogue data generation framework, and the framework comprises a synthetic primitive library which is used for storing one or more structured synthetic primitives, each synthetic primitive packages a complete closed-loop logic of a dialogue sub-task, and at least comprises four sub-modules which are mutually coupled; the expert guide system is used for receiving high-level task requirements, dynamically retrieving, combining and arranging matched synthetic primitives from the synthetic primitive library, and forming an executable primitive flow; and the modular generation engine is used for instantiating each synthetic primitive according to the primitive flow, driving a closed-loop process of generation-evaluation-reflection-strategy execution-regeneration, adding the dialogue fragments passing the quality evaluation to the global context, and generating multi-round dialogue data. According to the method, multi-round dialogue data which is coherent in context, consistent in logic, clear in task orientation, accurate in domain knowledge and capable of meeting business compliance requirements is automatically generated based on the structured synthesis primitives.

A system comprising a hierarchy of language model-based subsystems has been created to deliver responses to end-users that are within the scope of their queries. A first subsystem, or “conversation tracker,” receives user queries and tracks conversations with users. The conversation tracker keeps track of the conversation state, which comprises the historical context of a conversation a user has with the system. The conversation tracker additionally rephrases user queries to include the relevant historical context identified from the conversation. The conversation tracker also analyzes each rephrased user query and determines a domain (e.g., knowledge or task domain) to which the rephrased query corresponds applies. The conversation tracker forwards the rephrased queries to one of a plurality of additional subsystems, or “task resolvers,” for generation of responses that are ultimately relayed to the user.

The invention provides a large language model jailbreak system and method based on context history migration. The system comprises a scene embedding module, a malicious instruction generation module, a multi-round history generation module, a security scoring module, a history migration module and an execution strategy module. According to the method, target model parameters do not need to be accessed or modified, and the cross-platform adaptation capability is achieved under the standard API interface; task-oriented structure division is adopted in module design, function responsibilities are clear, maintenance and expansion are facilitated, and the universality and efficiency of the jail break attack method are greatly improved; during operation of the system, the number of calling times of the external model is small, the computing resource overhead is low, the output response is natural and coherent, a black box model jailbreak foundation platform with high success rate and low interference is built, and a powerful support foundation is provided for subsequent construction of a more complex instruction guidance system.

The invention relates to the field of neurological rehabilitation medicine, and particularly discloses application of task-oriented training in promoting functional recovery after spinal cord injury (SCI). The invention is based on the discovery that task-oriented training (such as single globule grasping training) can activate an NRG1 / ErbB4 signaling pathway by up-regulating the phosphorylation level of neuro-regulatory protein-1 (NRG1) and a receptor ErbB4 of the NRG1, thereby promoting regenerative cortical spinal bundle (CST) axons and spinal motor neurons to form functional excitatory synapses and inducing the conversion of synaptic phenotypes to excitatory. The invention provides a treatment method including task-oriented training and a combined strategy for combined application of the treatment method and a treatment method for activating endogenous axon regeneration capacity (such as by inhibiting PTEN expression), the treatment method is used for improving upper limb movement dysfunction after spinal cord injury, particularly after neck injury, and a new scheme is provided for clinical treatment.

The invention provides a task-oriented positioning method and system for sequential positioning of a 3D data set. Selecting a task execution scene 3D environment, screening a 3D environment scene graph, adding object subtitles to expand object node semantic attributes, and obtaining object node semantic attribute expansion information; according to object node semantic attribute extension information, a scene graph automatic generation pipeline is constructed, and multiple groups of daily tasks are automatically created; verifying the adaptability of each group of daily tasks to the task execution scene environment, and constructing a 3D sequential positioning verification data set; and performing 3D sequential positioning verification data set benchmark testing on the various 3D visual positioning models through the 3D sequential positioning verification data set, marking 3D sequential positioning verification data set positioning defects of the 3D visual positioning models, and performing training optimization on the various 3D visual positioning models.

The application relates to a kind of unsupervised domain adaptive semantic segmentation methods of thyroid ultrasound image based on feature decoupling, including feature decoupling and image translation two parts, the feature decoupling and image translation utilize feature constraint module and domain label coding module to carry out semantic segmentation.The feature constraint module of the application can improve the decoupling mode of structure-texture, based on the maximum entropy principle from the angle of task correlation, will domain invariant feature be expanded into task-oriented feature with strong task correlation, so that task-oriented feature contains rich detail information, improves the segmentation accuracy of nodule.Domain label coding module can learn the hidden feature representation of domain bias using domain label containing domain-specific information, and combine hidden features with task-oriented features to improve the stability of detail features, and then improve the domain invariance of task-oriented features, so that the segmentation model can more accurately distinguish benign and malignant nodules.

Embodiments described herein provide safe policy improvement (SPI) in a batch reinforcement learning framework for a task-oriented dialogue. Specifically, a batch reinforcement learning framework for dialogue policy learning is provided, which improves the performance of the dialogue and learns to shape a reward that reasons the invention behind human response rather than just imitating the human demonstration.

Provided is a task-oriented dialogue implementation method relating to artificial intelligence fields such as deep learning, large language models, natural language processing and intelligent agents, which can be applied to intelligent interaction scenarios such as intelligent customer service, intelligent outbound calling, and intelligent marketing. The task-oriented dialogue implementation method may include: acquiring a question to be answered; generate an answer corresponding to the question by using a task-oriented dialogue model, the answer is generated by the task-oriented dialogue model according to model configuration information, the model configuration information includes a dialogue flow corresponding to the task-oriented dialogue model, and the dialogue flow is written in text form.